Skip to content

Design and implement metrics for assessing input data quality and logic for approximating raw option prices #2

Description

@plkn

The provided test data is highly diverse, covering multiple markets and a wide range of underlying assets. Within each market, the characteristics of the available options data vary significantly: some underlyings have dense strike grids with tight spreads and high liquidity, while others have only a single expiration, a limited number of strikes, and very low liquidity.

The quality of the input options data has a strong impact on the performance of naive volatility-curve-building algorithms. For high-quality data (the happy path), even simple approaches may produce reasonable results. However, when working with sparse or illiquid data, these algorithms tend to produce unreliable volatility curves, reflecting the limitations of the input data itself.

The goal of this task is to design and implement metrics that evaluate the quality of the input options data. These metrics will allow the algorithm to adapt its behavior depending on the characteristics of the available data.

For low-quality datasets, where option quotes are sparse and bid-ask spreads are wide, directly approximating the volatility surface may be ineffective. In such cases, the proposed approach is to first approximate the observed option prices in price space and only then transform the resulting approximation into volatility space.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions